In-memory Computing and Sports Betting Sam Lawrence CTO at FSB - - PowerPoint PPT Presentation

In-memory Computing and Sports Betting

Sam Lawrence CTO at FSB Technology

Keynote Abstract

◼ Competition, innovation, regulation and improved data collection have grown the complexity of Sports Betting

systems at an unprecedented rate. Coupled with a massive growth in transaction volumes, Sports Betting provides a great case study for considering the relationship between technological advance and business need.

◼ This presentation is a personal look at the technical challenges the industry has faced, the solutions implemented

by FSB Tech and some best guesses of likely crunch points in the future.

Introduction - Sam Lawrence

◼ CTO and co-founder at FSB Technology,

previously CTO at GAN

◼ 18 years in gaming industry ◼ Started programming as 9 year old on Sinclair

ZX81

◼ Still fairly “hands-on”, but happily more hazy on

the details as the company grows

◼ Appeared on the cover of “Your Computer” in

1985

Presentation Summary

• 1. Sports Betting Technology - history and challenges
• 2. FSB’s approach and experience
• 3. In-Memory Computing - what I’ve learnt
• 4. Wrapping up

◼ Basic premise, money in a hat, winners share

based on race or match result

◼ Parimutuel or Tote betting ◼ Early C. 20th used mechanical solutions ◼ Fixed odds provide more betting options without

liquidity problems

◼ Online has transformed the industry - more bets

taken, more sports covered, more ways to bet

• 1. Sports Betting Technology - history and challenges

A blend of maths, finance, sport and technology

The world’s first parallel automatic totalisator - Ellerslie Racecourse New Zealand 1913 (http://rutherfordjournal.org/article020109.html)

Fixed odds betting needs accurate prediction models

◼ 1 / probability … but with margin or “overround” ◼ Complex Models, controlled by key parameters ◼ Faster / richer input data

◼ live sports data ◼ customer activity ◼ other bookmakers - avoiding arbitrage ◼ human traders

◼ Computing has given more techniques, but hasn't

changed much of the underlying maths

From the “Poisson distribution” page on Wikipedia

Challenges for building a Sports Betting Platform

◼ Volatility

◼ one-off sporting events ◼ Saturday afternoon

◼ Regulation

◼ protecting customers ◼ data protection ◼ spotting money laundering ◼ maintaining integrity in sport

◼ New Products

◼ e.g. Cash Out and Request a Bet

◼ Competition

◼ Welcome offers ◼ Odds Comparison

Picture of an automatic transmission valve body

Why is this relevant?

◼ Rapid change created opportunity ◼ Also presented many problems to solve ◼ Same problems faced by other industries ◼ Still need to innovate, not just react ◼ Because it’s the World Cup

• 2. FSB’s Approach and Experience

◼ Provides sports betting as a service

◼ White label websites ◼ Bespoke implementations

◼ Developer and Operator ◼ Founded in 2007, originally focused on

providing Fantasy Sports

◼ Launched first fully responsive site in 2014 ◼ Now run 36 branded sites in 11 different

countries in 4 continents

Simple beginnings 2008 - 2010

◼ Proof of concept platform ◼ PostgreSQL database ◼ GlassFish application server ◼ Lots of JMS ◼ Choice of open-source technologies with route to

enterprise support

◼ Caching and distribution of data with JPA

Glassfish EJB / JMS / JPA PostgreSQL

Growing pains and patches c. 2010 - 2014

◼ OpenJMS replaced with ActiveMQ ◼ Scope of JPA restricted - buggy and slow at high

volumes

◼ Caching of sports data via memcached ◼ MongoDB for document based data structures ◼ Added ESB with karaf / camel for external data

interfacing

Glassfish EJB / JMS / JPA MongoDB PostgreSQL Karaf + Camel

Memcached

Ready for growth c. 2014 - present

◼ GridGain caching and processing bespoke data

◼ sports data - more reads than writes ◼ fast risk and liability calculations

◼ GridGain for distributed locks ◼ PostgreSQL still “master” - final safeguard for data

integrity

◼ Scale horizontally adding VMs ◼ Now run 9 different instance of platform

Wildfly / ActiveMQ EJB / JMS / JPA MyBatis MongoDB PostgreSQL Karaf + Camel

GridGain

Where does FSB’s Technology go next?

◼ Platform stable and scalable - much new

development driven by customers

◼ The need to keep innovating

◼ Entrepreneur’s / CTO’s mindset? ◼ But how to measure value?

◼ Bigger customers want more “Enterprise”

guarantees

◼ Security, DR, minimal downtime

Growing a business takes patience

• 3. In-Memory Computing - what I’ve learnt

◼ The key drivers

◼ Increase performance ◼ Handle richer / faster data ◼ Allow horizontal scaling ◼ Redundancy

◼ Relationship with storage

◼ Faster / cheaper ◼ Data that needs to be kept ◼ Data that can’t be kept too long ◼ Persistent memory

What questions did I ask when looking for a solution?

◼ Performance ◼ Redundancy ◼ Scale with needs / volume ◼ Cost / benefits ◼ Community ◼ Support ◼ Development

What about the developers?

◼ Okay, you want to do In Memory computing ◼ Do you need a different mindset? ◼ What framework should you use?

◼ ...Events ◼ ...Streams ◼ ...Services ◼ ...Big data

◼ How do you develop?

◼ Use DevOps tools - ansible, VirtualBox

◼ How do you test?

◼ Concurrency issues are hard!

John Backus 1978 - “Can Programming Be Liberated from the von Neumann Style? A Functional Style and Its Algebra of Programs”

• 4. Wrapping up

◼ Sports Betting is an exciting industry for tech ◼ In-Memory computing was essential to enable scale ◼ Relieved to have a firm foundation, but still many problems to solve ◼ Excited about the future

Thank you!

◼ Sam Lawrence ◼ CTO at FSB Technology ◼ sam@fsbtech.com